Dual objective bandit for best channel selection in hybrid band wireless systems

This paper manipulates a creative online learning solution for optimal band/channel assignment in hybrid radio frequency and visible light communication (RF/VLC) systems. Typically, the multiband transmitter (Tx) has no prior information about different channel characteristics including their payoffs (achievable data rate) and energy consumption outcome. Practically, Tx has to choose the best reward arm/band with the lowest energy consumption to prolong its limited battery capacity. Hence, we envision a lightweight cost-aware multi-armed bandit (CA-MAB) as a proper realistic solution to the cumbersome and slowly convergent ordinary band assignment methods, where the transmitter/player intends not only to maximize his cumulative payoff (achievable data rate) but also to mitigate his cost (battery consumption due to the utilized band). Therefore, we propose a dual objective MAB scheme to manage such problem intelligently. Numerical simulations indicate that proposed method outperforms naive MAB versions, including Thompson sampling (TS), Upper Confidence bound (UCB), and traditional hybrid band selection (HBA) approaches, correspondingly. Especially, our proposed algorithm delivers 99% of the optimal performance concerning energy consumption, achievable data rate, and convergence speed.